Dialogue datasets are crucial for deep learning-based task-oriented dialogue system research. While numerous English language multi-domain task-oriented dialogue datasets have been developed and contributed to significant advancements in task-oriented dialogue systems, such a dataset does not exist in Japanese, and research in this area is limited compared to that in English. In this study, towards the advancement of research and development of task-oriented dialogue systems in Japanese, we constructed JMultiWOZ, the first Japanese language large-scale multi-domain task-oriented dialogue dataset. Using JMultiWOZ, we evaluated the dialogue state tracking and response generation capabilities of the state-of-the-art methods on the existing major English benchmark dataset MultiWOZ2.2 and the latest large language model (LLM)-based methods. Our evaluation results demonstrated that JMultiWOZ provides a benchmark that is on par with MultiWOZ2.2. In addition, through evaluation experiments of interactive dialogues with the models and human participants, we identified limitations in the task completion capabilities of LLMs in Japanese.
Empathetic response generation endeavors to empower dialogue systems to perceive speakers' emotions and generate empathetic responses accordingly. Psychological research demonstrates that emotion, as an essential factor in empathy, encompasses trait emotions, which are static and context-independent, and state emotions, which are dynamic and context-dependent. However, previous studies treat them in isolation, leading to insufficient emotional perception of the context, and subsequently, less effective empathetic expression. To address this problem, we propose Combining Trait and State emotions for Empathetic Response Model (CTSM). Specifically, to sufficiently perceive emotions in dialogue, we first construct and encode trait and state emotion embeddings, and then we further enhance emotional perception capability through an emotion guidance module that guides emotion representation. In addition, we propose a cross-contrastive learning decoder to enhance the model's empathetic expression capability by aligning trait and state emotions between generated responses and contexts. Both automatic and manual evaluation results demonstrate that CTSM outperforms state-of-the-art baselines and can generate more empathetic responses. Our code is available at https://github.com/wangyufeng-empty/CTSM
The proliferation of online misinformation has posed significant threats to public interest. While numerous online users actively participate in the combat against misinformation, many of such responses can be characterized by the lack of politeness and supporting facts. As a solution, text generation approaches are proposed to automatically produce counter-misinformation responses. Nevertheless, existing methods are often trained end-to-end without leveraging external knowledge, resulting in subpar text quality and excessively repetitive responses. In this paper, we propose retrieval augmented response generation for online misinformation (RARG), which collects supporting evidence from scientific sources and generates counter-misinformation responses based on the evidences. In particular, our RARG consists of two stages: (1) evidence collection, where we design a retrieval pipeline to retrieve and rerank evidence documents using a database comprising over 1M academic articles; (2) response generation, in which we align large language models (LLMs) to generate evidence-based responses via reinforcement learning from human feedback (RLHF). We propose a reward function to maximize the utilization of the retrieved evidence while maintaining the quality of the generated text, which yields polite and factual responses that clearly refutes misinformation. To demonstrate the effectiveness of our method, we study the case of COVID-19 and perform extensive experiments with both in- and cross-domain datasets, where RARG consistently outperforms baselines by generating high-quality counter-misinformation responses.
Large language models (LLMs) have shown tremendous success in following user instructions and generating helpful responses. Nevertheless, their robustness is still far from optimal, as they may generate significantly inconsistent responses due to minor changes in the verbalized instructions. Recent literature has explored this inconsistency issue, highlighting the importance of continued improvement in the robustness of response generation. However, systematic analysis and solutions are still lacking. In this paper, we quantitatively define the inconsistency problem and propose a two-stage training framework consisting of instruction-augmented supervised fine-tuning and consistency alignment training. The first stage helps a model generalize on following instructions via similar instruction augmentations. In the second stage, we improve the diversity and help the model understand which responses are more aligned with human expectations by differentiating subtle differences in similar responses. The training process is accomplished by self-rewards inferred from the trained model at the first stage without referring to external human preference resources. We conduct extensive experiments on recent publicly available LLMs on instruction-following tasks and demonstrate the effectiveness of our training framework.
Mitigating the generation of contradictory responses poses a substantial challenge in dialogue response generation. The quality and quantity of available contradictory response data play a vital role in suppressing these contradictions, offering two significant benefits. First, having access to large contradiction data enables a comprehensive examination of their characteristics. Second, data-driven methods to mitigate contradictions may be enhanced with large-scale contradiction data for training. Nevertheless, no attempt has been made to build an extensive collection of model-generated contradictory responses. In this paper, we build a large dataset of response generation models' contradictions for the first time. Then, we acquire valuable insights into the characteristics of model-generated contradictions through an extensive analysis of the collected responses. Lastly, we also demonstrate how this dataset substantially enhances the performance of data-driven contradiction suppression methods.
Large Language Models (LLMs) exhibit impressive capabilities but also present risks such as biased content generation and privacy issues. One of the current alignment techniques includes principle-driven integration, but it faces challenges arising from the imprecision of manually crafted rules and inadequate risk perception in models without safety training. To address these, we introduce Guide-Align, a two-stage approach. Initially, a safety-trained model identifies potential risks and formulates specific guidelines for various inputs, establishing a comprehensive library of guidelines and a model for input-guidelines retrieval. Subsequently, the retrieval model correlates new inputs with relevant guidelines, which guide LLMs in response generation to ensure safe and high-quality outputs, thereby aligning with human values. An additional optional stage involves fine-tuning a model with well-aligned datasets generated through the process implemented in the second stage. Our method customizes guidelines to accommodate diverse inputs, thereby enhancing the fine-grainedness and comprehensiveness of the guideline library. Furthermore, it incorporates safety expertise from a safety-trained LLM through a lightweight retrieval model. We evaluate our approach on three benchmarks, demonstrating significant improvements in LLM security and quality. Notably, our fine-tuned model, Labrador, even at 13 billion parameters, outperforms GPT-3.5-turbo and surpasses GPT-4 in alignment capabilities.
Depression-diagnosis-oriented chat aims to guide patients in self-expression to collect key symptoms for depression detection. Recent work focuses on combining task-oriented dialogue and chitchat to simulate the interview-based depression diagnosis. Whereas, these methods can not well capture the changing information, feelings, or symptoms of the patient during dialogues. Moreover, no explicit framework has been explored to guide the dialogue, which results in some useless communications that affect the experience. In this paper, we propose to integrate Psychological State Tracking (POST) within the large language model (LLM) to explicitly guide depression-diagnosis-oriented chat. Specifically, the state is adapted from a psychological theoretical model, which consists of four components, namely Stage, Information, Summary and Next. We fine-tune an LLM model to generate the dynamic psychological state, which is further used to assist response generation at each turn to simulate the psychiatrist. Experimental results on the existing benchmark show that our proposed method boosts the performance of all subtasks in depression-diagnosis-oriented chat.
Existing neural response generation models have achieved impressive improvements for two-party conversations, which assume that utterances are sequentially organized. However, many real-world dialogues involve multiple interlocutors and the structure of conversational context is much more complex, e.g. utterances from different interlocutors can occur "in parallel". Facing this challenge, there are works trying to model the relations among utterances or interlocutors to facilitate response generation with clearer context. Nonetheless, these methods rely heavily on such relations and all assume that these are given beforehand, which is impractical and hinders the generality of such methods. In this work, we propose to automatically infer the relations via relational thinking on subtle clues inside the conversation context without any human label, and leverage these relations to guide the neural response generation. Specifically, we first apply a deep graph random process to fully consider all possible relations among utterances in the conversational context. Then the inferred relation graphs are integrated with a variational auto-encoder framework to train a GAN for structure-aware response generation. Experimental results on the Ubuntu Internet Relay Chat (IRC) channel benchmark and the most recent Movie Dialogues show that our method outperforms various baseline models for multi-party response generation.
While Reinforcement Learning from Human Feedback (RLHF) effectively aligns pretrained Large Language and Vision-Language Models (LLMs, and VLMs) with human preferences, its computational cost and complexity hamper its wider adoption. To alleviate some of the computational burden of fine-tuning, parameter efficient methods, like LoRA were introduced. In this work, we empirically evaluate the setup of Parameter Efficient Reinforcement Learning from Human Feedback (PE-RLHF) that leverages LoRA fine-tuning for Reward Modeling, and Reinforcement Learning. We benchmark the PE-RLHF setup on six diverse datasets spanning summarization, harmless/helpful response generation, UI automation, and visual question answering in terms of effectiveness of the trained models, and the training resources required. Our findings show, for the first time, that PE-RLHF achieves comparable performance to RLHF, while significantly reducing training time (up to 90% faster for reward models, and 30% faster for RL), and memory footprint (up to 50% reduction for reward models, and 27% for RL). We provide comprehensive ablations across LoRA ranks, and model sizes for both reward modeling and reinforcement learning. By mitigating the computational burden associated with RLHF, we push for a broader adoption of PE-RLHF as an alignment technique for LLMs and VLMs.
Large vision-language models (LVLMs), designed to interpret and respond to human instructions, occasionally generate hallucinated or harmful content due to inappropriate instructions. This study uses linear probing to shed light on the hidden knowledge at the output layers of LVLMs. We demonstrate that the logit distributions of the first tokens contain sufficient information to determine whether to respond to the instructions, including recognizing unanswerable visual questions, defending against jailbreaking attacks, and identifying deceptive questions. Such hidden knowledge is gradually lost in logits of subsequent tokens during response generation. Then, we illustrate a simple decoding strategy at the generation of the first token, effectively improving the generated content. In experiments, we find a few interesting insights: First, the CLIP model already contains a strong signal for solving these tasks, which indicates potential bias in the existing datasets. Second, we observe performance improvement by utilizing the first logit distributions on three additional tasks, including indicating uncertainty in math solving, mitigating hallucination, and image classification. Last, with the same training data, simply finetuning LVLMs improves models' performance but is still inferior to linear probing on these tasks.
Conversational systems are crucial for human-computer interaction, managing complex dialogues by identifying threads and prioritising responses. This is especially vital in multi-party conversations, where precise identification of threads and strategic response prioritisation ensure efficient dialogue management. To address these challenges an end-to-end model that identifies threads and prioritises their response generation based on the importance was developed, involving a systematic decomposition of the problem into discrete components - thread detection, prioritisation, and performance optimisation which was meticulously analysed and optimised. These refined components seamlessly integrate into a unified framework, in conversational systems. Llama2 7b is used due to its high level of generalisation but the system can be updated with any open source Large Language Model(LLM). The computational capabilities of the Llama2 model was augmented by using fine tuning methods and strategic prompting techniques to optimise the model's performance, reducing computational time and increasing the accuracy of the model. The model achieves up to 10x speed improvement, while generating more coherent results compared to existing models.
For a human-like chatbot, constructing a long-term memory is crucial. However, current large language models often lack this capability, leading to instances of missing important user information or redundantly asking for the same information, thereby diminishing conversation quality. To effectively construct memory, it is crucial to seamlessly connect past and present information, while also possessing the ability to forget obstructive information. To address these challenges, we propose CREEM, a novel memory system for long-term conversation. Improving upon existing approaches that construct memory based solely on current sessions, CREEM blends past memories during memory formation. Additionally, we introduce a refining process to handle redundant or outdated information. Unlike traditional paradigms, we view responding and memory construction as inseparable tasks. The blending process, which creates new memories, also serves as a reasoning step for response generation by informing the connection between past and present. Through evaluation, we demonstrate that CREEM enhances both memory and response qualities in multi-session personalized dialogues.
Knowledge-based, open-domain dialogue generation aims to build chit-chat systems that talk to humans using mined support knowledge. Many types and sources of knowledge have previously been shown to be useful as support knowledge. Even in the era of large language models, response generation grounded in knowledge retrieved from additional up-to-date sources remains a practically important approach. While prior work using single-source knowledge has shown a clear positive correlation between the performances of knowledge selection and response generation, there are no existing multi-source datasets for evaluating support knowledge retrieval. Further, prior work has assumed that the knowledge sources available at test time are the same as during training. This unrealistic assumption unnecessarily handicaps models, as new knowledge sources can become available after a model is trained. In this paper, we present a high-quality benchmark named multi-source Wizard of Wikipedia (Ms.WoW) for evaluating multi-source dialogue knowledge selection and response generation. Unlike existing datasets, it contains clean support knowledge, grounded at the utterance level and partitioned into multiple knowledge sources. We further propose a new challenge, dialogue knowledge plug-and-play, which aims to test an already trained dialogue model on using new support knowledge from previously unseen sources in a zero-shot fashion.
It is often desirable for Large Language Models (LLMs) to capture multiple objectives when providing a response. In document-grounded response generation, for example, agent responses are expected to be relevant to a user's query while also being grounded in a given document. In this paper, we introduce Proxy Metric-based Self-Refinement (ProMiSe), which enables an LLM to refine its own initial response along key dimensions of quality guided by external metrics feedback, yielding an overall better final response. ProMiSe leverages feedback on response quality through principle-specific proxy metrics, and iteratively refines its response one principle at a time. We apply ProMiSe to open source language models Flan-T5-XXL and Llama-2-13B-Chat, to evaluate its performance on document-grounded question answering datasets, MultiDoc2Dial and QuAC, demonstrating that self-refinement improves response quality. We further show that fine-tuning Llama-2-13B-Chat on the synthetic dialogue data generated by ProMiSe yields significant performance improvements over the zero-shot baseline as well as a supervised fine-tuned model on human annotated data.
Self-consistency has emerged as a powerful method for improving the accuracy of short answers generated by large language models. As previously defined, it only concerns the accuracy of a final answer parsed from generated text. In this work, we extend the idea to open response generation, by integrating voting into the decoding method. Each output sentence is selected from among multiple samples, conditioning on the previous selections, based on a simple token overlap score. We compare this "Sample & Select" method to greedy decoding, beam search, nucleus sampling, and the recently introduced hallucination avoiding decoders of DoLA, P-CRR, and S-CRR. We show that Sample & Select improves factuality by a 30% relative margin against these decoders in NLI-based evaluation on the subsets of CNN/DM and XSum used in the FRANK benchmark, while maintaining comparable ROUGE-1 F1 scores against reference summaries. We collect human verifications of the generated summaries, confirming the factual superiority of our method.
We present the second version of the Open Assistant Toolkit (OAT-v2), an open-source task-oriented conversational system for composing generative neural models. OAT-v2 is a scalable and flexible assistant platform supporting multiple domains and modalities of user interaction. It splits processing a user utterance into modular system components, including submodules such as action code generation, multimodal content retrieval, and knowledge-augmented response generation. Developed over multiple years of the Alexa TaskBot challenge, OAT-v2 is a proven system that enables scalable and robust experimentation in experimental and real-world deployment. OAT-v2 provides open models and software for research and commercial applications to enable the future of multimodal virtual assistants across diverse applications and types of rich interaction.
Empathetic response generation aims to comprehend the cognitive and emotional states in dialogue utterances and generate proper responses. Psychological theories posit that comprehending emotional and cognitive states necessitates iteratively capturing and understanding associated words across dialogue utterances. However, existing approaches regard dialogue utterances as either a long sequence or independent utterances for comprehension, which are prone to overlook the associated words between them. To address this issue, we propose an Iterative Associative Memory Model (IAMM) for empathetic response generation. Specifically, we employ a novel second-order interaction attention mechanism to iteratively capture vital associated words between dialogue utterances and situations, dialogue history, and a memory module (for storing associated words), thereby accurately and nuancedly comprehending the utterances. We conduct experiments on the Empathetic-Dialogue dataset. Both automatic and human evaluations validate the efficacy of the model. Variant experiments on LLMs also demonstrate that attending to associated words improves empathetic comprehension and expression.
Adversarial misuse, particularly through `jailbreaking' that circumvents a model's safety and ethical protocols, poses a significant challenge for Large Language Models (LLMs). This paper delves into the mechanisms behind such successful attacks, introducing a hypothesis for the safety mechanism of aligned LLMs: intent security recognition followed by response generation. Grounded in this hypothesis, we propose CodeChameleon, a novel jailbreak framework based on personalized encryption tactics. To elude the intent security recognition phase, we reformulate tasks into a code completion format, enabling users to encrypt queries using personalized encryption functions. To guarantee response generation functionality, we embed a decryption function within the instructions, which allows the LLM to decrypt and execute the encrypted queries successfully. We conduct extensive experiments on 7 LLMs, achieving state-of-the-art average Attack Success Rate (ASR). Remarkably, our method achieves an 86.6\% ASR on GPT-4-1106.
Empathetic response generation aims to generate empathetic responses by understanding the speaker's emotional feelings from the language of dialogue. Recent methods capture emotional words in the language of communicators and construct them as static vectors to perceive nuanced emotions. However, linguistic research has shown that emotional words in language are dynamic and have correlations with other grammar semantic roles, i.e., words with semantic meanings, in grammar. Previous methods overlook these two characteristics, which easily lead to misunderstandings of emotions and neglect of key semantics. To address this issue, we propose a dynamical Emotion-Semantic Correlation Model (ESCM) for empathetic dialogue generation tasks. ESCM constructs dynamic emotion-semantic vectors through the interaction of context and emotions. We introduce dependency trees to reflect the correlations between emotions and semantics. Based on dynamic emotion-semantic vectors and dependency trees, we propose a dynamic correlation graph convolutional network to guide the model in learning context meanings in dialogue and generating empathetic responses. Experimental results on the EMPATHETIC-DIALOGUES dataset show that ESCM understands semantics and emotions more accurately and expresses fluent and informative empathetic responses. Our analysis results also indicate that the correlations between emotions and semantics are frequently used in dialogues, which is of great significance for empathetic perception and expression.
In this study, we tackle a growing concern around the safety and ethical use of large language models (LLMs). Despite their potential, these models can be tricked into producing harmful or unethical content through various sophisticated methods, including 'jailbreaking' techniques and targeted manipulation. Our work zeroes in on a specific issue: to what extent LLMs can be led astray by asking them to generate responses that are instruction-centric such as a pseudocode, a program or a software snippet as opposed to vanilla text. To investigate this question, we introduce TechHazardQA, a dataset containing complex queries which should be answered in both text and instruction-centric formats (e.g., pseudocodes), aimed at identifying triggers for unethical responses. We query a series of LLMs -- Llama-2-13b, Llama-2-7b, Mistral-V2 and Mistral 8X7B -- and ask them to generate both text and instruction-centric responses. For evaluation we report the harmfulness score metric as well as judgements from GPT-4 and humans. Overall, we observe that asking LLMs to produce instruction-centric responses enhances the unethical response generation by ~2-38% across the models. As an additional objective, we investigate the impact of model editing using the ROME technique, which further increases the propensity for generating undesirable content. In particular, asking edited LLMs to generate instruction-centric responses further increases the unethical response generation by ~3-16% across the different models.